Research on an improved PSO algorithm with dual self-adaptation and dual variation

• Published in: 2022 IEEE International Conference on Mechatronics and Automation (ICMA) pp. 646 - 650
• Main Authors: Chen, Mingjie, Wang, Yuchao, Li, Peng, Fu, Huixuan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07.08.2022
• Subjects: Behavioral sciences; Birds; function optimization; improved PSO algorithm; Mechatronics; Nonlinear self-adaptation of inertia weight and linear self-adaptation of learning factors; Particle swarm optimization; Simulation; Sociology; Statistics; variation from motion direction and initial value of particles
• ISSN: 2152-744X
• DOI: 10.1109/ICMA54519.2022.9856223

Particle swarm optimization (PSO) algorithm is a global optimization evolutionary algorithm which simulates the foraging behavior of birds. In view of the problems of premature improved particle swarm optimization algorithm with dual self-convergence and prematurity of standard PSO algorithm, an adaptation and dual variation is proposed, to overcome the shortcomings of the existing improved PSO algorithm with linear inertia weight and motion direction variation. A nonlinear self-adaptive improvement of inertia weight and a linear self-adaptive improvement of learning factor are presented to improve the scope and precision of the optimal search for the balanced population, to strengthen the global search ability in the initial stage of optimization and to enhance the local search ability in the later stage of optimization. Moreover, in the later period of motion, the variation from the motion directions of some particles and the variation from initial values of a few particles are used to increase the diversity of the population and enlarge the searching range of the particles, so as to prevent the particles from falling into the local optimal solution. The improved PSO algorithm with dual self-adaptation and dual variation is applied to function optimization. The simulation results show that the algorithm can effectively improve the convergence rate and precision of the algorithm, and avoid premature convergence.